v0.4.0 v0.4.1
VectifyAI/PageIndexOct 12, 2026 → Oct 16, 20264 days between releases
What's New
Flash engine gains optional LLM node summaries for flexible indexing.
New Features
- LiteLLM-routed providers for flexible model connections.
- Explicit `PageIndexLocalClient` and `PageIndexCloudClient` constructors.
- MCP tool contract support with specific config bundles.
- Native support for multiple streaming protocols (Responses/Messages).
Improvements
- Flash engine now allows optional LLM node summaries alongside layout stats.
- Enhanced model and connection knobs for better configuration control.
- Unified API now explicitly separates local and cloud client constructors.
Upgrade Notes
The unified client API remains unchanged, though explicit constructors for `PageIndexLocalClient` and `PageIndexCloudClient` are now available. Configuration options have expanded to include `index_model`, `chat_model`, and LiteLLM-routed providers. No breaking changes to the core interface are expected.
The Two Releases
v0.4.0
Oct 12, 2026· 9 days ago
PageIndex v0.4.0 introduces the Flash engine to drastically reduce indexing costs and time by using layout stats instead of LLMs for tree generation. The SDK now provides a unified client interface for both local and cloud operations, supports custom chat models for cloud documents, and includes comprehensive agent integration tools.
Full release notesv0.4.1
Oct 16, 2026· 6 days ago
This release introduces the Flash engine for faster and cheaper indexing by using layout statistics instead of LLMs for structure generation. It provides a unified client interface that works seamlessly in both local and cloud modes, with no server, vector DB, or API key required for local use.
Full release notes